On using the wisdom of the crowd principles in classification, Application on breast cancer diagnosis and prognosis.

Breast cancer diagnosis and prognosis are an oblique processes, where errors can be fatal, it is done by experts only. Therefore, researchers are using the promising potentials of classification algorithms to detect malignant and benign tumours. Classification techniques vary widely, from individual classifiers such as rules, trees and functions to ensemble classifiers that combine serval classification algorithms. In this paper, we examine the use of wisdom of crowds' principles in classification of breast cancer for diagnosis and prognosis. We use four well-known datasets and run a collection of 53 algorithms combined with majority voting to simulate the wisdom of crowds. Furthermore, we report the results obtained from all of 53 algorithms executed individually on all four datasets. Therefore, this article can be perceived as a review for these classification methods as well. Moreover, we compare the results obtained from applying majority voting using the best five classifiers, to those obtained by applying the wisdom of the crowds' principles. The results will show that outcome of having too many classifiers reduces the accuracy dramatically. Consequently, it is not advisable to use more than five strong classifiers.